WHITE PAPER. Reliability The ShoreTel Way

Transcription

1 WHITE PAPER Reliability The Way

2 Table of Contents 1. Introduction s Distributed Architecture N+1 Redundancy Highly Reliable Components Remote Survivability Reliable Remote Office Applications Application Server Solutions Centralized Services Distributed Services Disaster Recovery Solutions Double-Take VMware Enterprise Contact Center Disaster Recovery Summary Figure 1. Distributed Architecture Unaffected by WAN Failure...4 Figure 2. Distributed Call Control Unaffected by WAN Failure...8 Figure 3. V Models Provide Voice Mail and Auto Attendant Features... 9 Figure 4. Typical Element Multi-Site Administration Figure 5. Distributed Applications Figure 6. Generic VMware Solution for Reliability The Way PAGE 2

3 1. Introduction Reliability is a critical component of any business communication system. understands this and provides one of the most robust and reliable unified communications (UC) solutions in the world. s architecture combines unprecedented reliability with a brilliantly simple design at a low total cost of ownership. designed a pure IP telephony solution from the start instead of having to convert an existing TDM product to a hybrid (TDM and IP) solution. This approach gives built-in mechanisms that quickly and cost effectively route communications over a customer WAN and deliver the necessary enterprise-level reliability. The UC system delivers industry-leading reliability that includes: A distributed architecture with an appliance approach for IP networks Multiple failover mechanisms which add resiliency and flexibility Centralized and/or distributed application services that fit any system architecture Comprehensive disaster recovery solutions that deliver enterprise caliber reliability Reliability The Way PAGE 3

4 2.. s Distributed Architecture The key component to s philosophy is a distributed architecture. uses the mantra of No single point of failure as a guiding principle in its communication system design. This philosophy drives the design principles that ensure the communication system offers high availability because any system down time correlates to loss of money; through either loss of revenue directly or increased system and maintenance costs. developed this highly differentiated architecture by combining the following features: N+1 redundancy instead of costly 1:1 redundancy A modular design that is superior to chassis, daughter boards or server designs High reliability components (flash media, RTOS, reduced number of components) that deliver a 60 year MTBF, which is approximately ten times higher than a typical PBX Robust survivability mechanisms (IP failover, dual LAN connections, power failure transfer, Contact center failover, trunk failover) Figure 1 gives an example of a distributed architecture for a medium-size firm that has three office locations and some remote/traveling workers. Users connect to their primary switch but have the ability to fail-over to alternate switches and servers across the network as needed. Director HOME OFFICE VPN Phone Call Manager PSTN /ITSP Analog Device HEADQUARTERS Headquarters Server (system administration, voic , auto attendant, work groups) Call Manager INTERNATIONAL BRANCH Voice Switches Contact Center Call Manager Distributed Voice Server (voic & auto attendant) Appliances (Conferencing & VPN Concentrator) IP Phone IP WAN IP Phone Contact Center Voice Switches Voice Switch Legacy Voic Call Manager for Mobile Call Manager for Desktop PSTN /ITSP Legacy Phones Legacy PBX REGIONAL OFFICE OFFICE ANYWHERE PSTN SINGLE-IMAGE ARCHITECTURE Figure 1. Distributed Architecture Unaffected by WAN Failure Reliability The Way PAGE 4

5 The IP network allows the distributed users (phones) many ways to connect to the voice switch(es) and maintain business continuity. 2.1 N+1 Redundancy Most IP PBXs are implemented with a central server (sometimes called a media gateway controller) that runs the call control to set up calls and provide telephony features. These systems use 1 to 1 (1:1) redundancy to provide percent (five-nines) availability. In contrast, uses a distributed call control model based upon an appliance approach, in which each voice switch provides call control for a limited number of phones (and PSTN lines) while working seamlessly with other appliances to support system users. A system is implemented by combining multiple voice switches that can provide full featured reliability throughout the network. By distributing the call control, voice switches can be made smaller, simpler and more reliable as each unit takes on only a portion of the overall task. s unique, distributed architecture allows the unit of redundancy to be much smaller. A UC system can be made redundant by the addition of an incremental unit, rather than duplicating the entire system. This is called n+1 redundancy and allows the business decision maker to decide the level of reliability (and cost) of the communication system. The business manager simply adds the number (n) of extra voice switches that they want. Under normal operation, the load is spread across all units. When one unit fails, the load is redistributed across the remaining units. If the system is constructed with extra capacity, there will be enough capacity for all users, and no system degradation. s n+1 redundancy is so much simpler and more cost-effective than the 1:1 redundancy used by other systems. For an n + 1 architecture, redundancy costs an extra 1/n. For a 1:1 architecture, redundancy costs 2n. There are some additional benefits to the use of n+1 redundancy. Firstly, as there are multiple units, the system remains resilient even after a failure, as there are still n units remaining. The system degrades with each subsequent failure, but remains functioning for some users. Secondly, the units can be geographically distributed, making disaster recovery easier and redundancy much more cost effective for smaller, remote sites. This technology also protects against WAN failures. A third benefit is that the system can scale without a forklift upgrade simply by adding phone switches. Reliability The Way PAGE 5

6 2.2 Highly Reliable Components units have a single board construction with very few hardware components and moving pieces. For instance, the UC system uses small power supplies, does not use cables or connectors within the chassis, and requires only a few fan units. All of these attributes make the system very reliable and are a direct consequence of s distributed architecture. In contrast, other vendors still typically use either a separate server or PC, classic chassis-based unit, or a horizontal daughter board based design in their IP PBX systems. These designs have the following inherent weaknesses: The server/pc has numerous cables, connectors, circuit boards and large power supplies. The classic chassis contains a backplane, numerous connectors, large power supplies and fans in addition to the large number of components on each circuit board. The daughter-board design eliminates the backplane but may require adding cables, while retaining the connectors, and a number of components on each circuit board. The main issue is that fans and large power supplies have a high probability of failure. The cables and connectors in these systems are frequently neglected in reliability computations, as it is difficult to assess mechanical reliability. But system administrators are familiar with having to re-seat cards and cables in these types of systems after moving them in order to get the equipment operating again. These are anecdotal observations, but reflect the actual problems of mechanical components in systems. Taking these reliability effects into account, along with the large component count in these configurations, their reliability is greatly reduced. Without redundancy, these systems typically deliver 99.9 percent (three-nines) availability rather than the target of percent. Flash memory is another differentiator for. By using this type of memory, we eliminate spinning media in the system. Spinning media has been shown to have a very short lifespan, on the order of two to three years, which dramatically decreases the mean time between failure (MTBF) of most systems. By contrast, systems deliver 60 years for an MTBF. 2.3 Remote Survivability Remote survivability mechanisms are another significant part of system reliability. Any IPbased system that intends to support percent reliability must support a capability to overcome the inherent unreliability of current IP networks. It is widely known that typical LAN and WAN network architectures are at best 99.9 percent reliable. Therefore, an IP communication system must address this unreliability with capabilities that overcome the weaknesses. Reliability The Way PAGE 6

7 To accomplish this goal, systems implement the following survivability mechanisms: IP Phone Failover: Each IP phone maintains a heartbeat with its controlling IP phone switch. Should the unit become unavailable, the IP phone will automatically reregister with the system and be assigned to another IP phone switch. IP phone switch Remote Failover: Multisite n+1 IP phone switch availability can be achieved with a single extra unit at the headquarters site that can automatically cover for a failed IP phone switch at a remote site. Spare units do not have to be maintained at each remote site (as in required in the 1:1 redundancy model), which is a significant cost savings as the number of sites grows, reducing the cost of redundancy nearly in half. Dual LAN connections: Allows for diverse networking in the event of a LAN failure. Backup Destination: When call control finds that an endpoint is unreachable, it attempts to use a configurable backup destination. This ensures that calls do not ring forever or go nowhere due to failures or incorrect configuration. Power Failure Transfer: Power transfer fail-over allows for fail-over of an analog station to an analog trunk (usually ports 8 and 9) in the event that there is a power failure to the equipment. Public Switched Telephone Network (PSTN) Failover: If the WAN becomes unavailable, the system will dial out on a trunk to the destination PSTN Direct Inward Dial (DID) number to reach the remote site. Failover Trunk Groups: Each user is a member of a user group that can be configured with a prioritized list of trunk groups; with least-cost routing determining which trunk gets used in any given instance. A failover trunk group can be configured in the event of network or hardware failure. All IP phone switch models have trunk interfaces that can be used for fallback. A small number of analog trunks can also be used for fallback and provide PSTN connectivity when digital trunks become unavailable. Contact Center Failover: The Enterprise Contact Center runs in a hot standby configuration where the secondary server can be distributed across the WAN. Configuration and reporting data is synchronized actively to the standby servers and in case of server, network or WAN outages all agents supervisors automatically fail over to the standby server ensuring business continuity. As mentioned earlier, LANs are less reliable primarily because they are implemented with multiple serially connected components. It is possible to achieve percent availability on the network by using a redundant aggregation switch that provides redundant paths to the IP PBX; which is commonly deployed for large and midsized customers. Smaller customers tend to forego redundancy and rely on the inherent reliability of the equipment and a quick service call in case of an outage content with a three-nines system and an expected downtime of eight hours a year. Reliability The Way PAGE 7

8 Reliable WANs are the major challenge in networking, because unlike LANs, they are not under the control of the individual enterprise, nor even the individual network provider. WAN reliability numbers are not generally available, but our experience suggests that WAN links are available for basic connectivity 99 percent to 99.9 percent of the time, with voice quality availability perhaps as low as 98 percent. The best of breed network providers promise percent availability when the WAN is implemented entirely by that single service provider. There are no known network providers with a percent service guarantee. When the WAN fails for IP-PBXs, the remote site loses contact with centralized call control the single brain of the system. Phones at the remote site are placed in survivable mode and use a fallback call control located in the site s gateway. This means that the remote site loses the benefit of the detailed call control rules and permissions contained only at the central call control location. Call routing at the survivable site loses higher business logic, call permissions, and restrictions, and often resorts to functionality comparable to nothing more than the equivalent of a basic home phone line. In addition, any higher level telephony applications and services are unavailable. This loss of features and functionality and fallback to nothing but PSTN access is often not acceptable for many businesses that rely on telephony features and applications to run their businesses. The situation is entirely different in the case of WAN failure. There is no change in call control as the local phone switch provides call control for the site and continues to do so in the case of WAN failure. The call control in the phone switch maintains a cached copy of the database and business logic (see Figure 2). The database is maintained at headquarters and change notifications are sent to all phone switches in the system, while individual units update their database caches. The remote site acts as a fully functional, independent phone system when the WAN is down, and then transparently rejoins the main system when the WAN becomes available. Remote Site Headquarters Remote Site Headquarters PSTN PSTN Voice Switch WAN X Notification Gateway WAN X Server Call Control Fallback Call Control Call Control ( The Brain ) Business Logic Database Business Logic Cache Database Distributed Call Control Centralized Call Control Figure 2. Distributed Call Control Unaffected by WAN Failure Reliability The Way PAGE 8

9 2.4 Reliable Remote Office Applications As mentioned in Section 2.3, when there is a WAN failure, IP communication systems must respond or else degradation/failure occurs. Besides the call-control requirements, there are application needs as well. The most critical of these applications at the remote site are: voic , auto attendant, hunt groups and in the case of, the desktop communication application Communicator. Remote Communicator solutions are addressed in Section 3.2. responds brilliantly to this situation by offering the following features: Built-in voic for remote switches Back-up auto attendant capability Workgroup to Hunt Group failover Contact Center failover For smaller sites, voice switches with integrated voic eliminate the need for a server. These are the V series switches 50V, 90V, and 90BRIV (illustrated in Figure 3 below) which provide voic /auto-attendant and flash memory storage. Figure 3. V Models Provide Voice Mail and Auto Attendant Features In the case of a WAN outage, the telephones will re-home to a redundant Voice Switch and voic /auto-attendant services will be provided by a server or V series switch. Voic messages for mailboxes on a different switch or server will be queued locally and forwarded upon connection restoration. If call control cannot reach an application server, each phone switch provides a backup auto-attendant function to notify the user that the destination is unavailable and offers to connect the user to another number to provide call answering. The autoattendant answers with a dial by number capability and the option to zero-out to the main menu. The Workgroup is commonly used as an informal contact center. In the event that the Workgroup service fails, calls can be sent to a hunt group to ensure that the call is answered. Reliability The Way PAGE 9

10 For Enterprise Contact Center, configuration and reporting data is synchronized actively to the standby servers. In the case of a server, network or WAN outage, all agents and/or supervisors automatically fail over to the standby server ensuring business continuity. Having standby servers ensures continuity in customer experience and minimizes any critical disruption to the support or revenue center. 3. Application Server Solutions has the ability to provide centralized and distributed services. This is one of the areas that differentiates from the competition. With a single image system, all voice switches are aware of each other and the stations assigned to each switch. In this architecture, the management is centralized and services (such as voic ) are typically centralized on a server, called the Headquarters (HQ). However, services like voic can also be distributed either on the telephone switches themselves or on other servers, call Distributed Voic Servers (DVS). The following sections discuss both centralized and distributed services. 3.1 Centralized Services s maintenance capabilities enable administrators to make major reconfigurations or upgrades and test them in as little as an hour. These same tasks typically take much longer on other IP phone systems, which lack s simple Director single administrative interface, and must be administered by experts from an error-prone command-line interface. owes this distinction to its simple design and architecture. Configuration changes are quick and easy because there is one single distributed system, regardless of the number of sites, and all changes local or remote are made from the Director console. Director is an Internet browser-based product that allows access to the equipment (voice switches, voic , auto-attendant, automatic call distributor, computer telephony applications and call detail records) from anywhere. recommends implementing the system with multiple servers to avoid a single point of failure. The headquarters server can be backed up with an additional server, bringing the number of servers to n+1, modeling the n+1 availability model of the voice switches. In the case of a server failure, the voice switch finds an alternative voic / auto-attendant server. The servers are arranged in a hierarchy maintained by the system administrator. When the voic /auto-attendant server assigned to a voice switch is unavailable, the voice switch analyzes its database until it finds a server and then uses it until its assigned voic /auto-attendant server becomes available. Competitive systems typically require separate administration of each of the major components of the system, imposing a heavy administrative burden as illustrated in Reliability The Way PAGE 10

11 Figure 4 below. On top of the complexity of a typical competitive multisite solution, all of the components at a remote site are a single point of failure, as 1:1 redundancy is too costly for small sites. The solution avoids these difficulties. Figure 4. Typical Element Multi-Site Administration Note that there are 18 elements to configure: six per system/site. Voic and desktop integration are replicated in order to achieve reliable applications. To stitch the separate phone systems together, dial plan and call route tables must be separately configured for system to system calling or long distance calling, adding cost and complexity to the solution. The Telephony Management Server ( Director) coordinates the voice switches and application services on each server. The Telephony Management Server service notifies all the voice switches and services that are managed by this server of configuration changes and passes configuration data to them. Telephony application programming interfaces for server and desktop applications are also provided by the Telephony Management Server, allowing every application access to every phone. It also provides the secret sauce that ties all the phones, voice switches, servers and services together into a single image system. 3.2 Distributed Services The approach to distributed services is to embed application services in its distributed architecture. A Distributed Voice Services (DVS) server can be located at each site. The system maintains its single administrative interface Director, and unlike competitive solutions, does not require duplication. Each server duplicates call control, voic , and desktop integration (the server side of the desktop application, Communicator). Reliability The Way PAGE 11

12 Director Server Telephony Voice Switch Management Management Call Control Server Interface Config. Call Control Data Site 1 Voice Mail Desktop Integration Server Config. Data Telephony Site 2 Management Server Voice Switch Call Control Call Control Voice Mail Config. Data Desktop Integration Config. Data Server Telephony Management Voice Switch Server Call Control Call Control Config. Voice Mail Site 3 Data Desktop Integration Config. Data Figure 5. Distributed Applications The architecture also allows additional instances of the configuration database to be installed within a system. This is typically implemented in a system with remote sites and a customer system that uses a DVS server. Changes made to the master database at the HQ location are automatically propagated to the remote sites. DVS servers can restart without requiring connectivity to the HQ location which eliminates the WAN as a functional and performance bottleneck. Reliability The Way PAGE 12

13 4. Disaster Recovery Solutions The system was designed for flexibility. In addition to the various options that can provide natively to customers, we have augmented our solution with solutions from third parties to deliver an unparalleled disaster recovery capability for the SMB and enterprise market. 4.1 Double-Take decided to enhance its own highly reliable solution with a third-party product from Double-Take. Double-Take s data replication technology combines continuous realtime backup and automatic failover capabilities for disaster recovery, high availability, and centralized backup. Double-Take goes beyond periodic backups to provide accessible and affordable data protection via data replication software. This ensures minimal data loss and enables immediate data recovery with failover from any disaster or system outage. Double-Take uses patented data replication and failover technology that continuously captures byte-level changes as they happen and replicates those changes to one or more target servers at any location, either locally or at a recovery site miles away. The Double-Take solution provides the following benefits: Centralized reporting and analysis Flexible bandwidth scheduling High availability and failover Integrates with existing backup solutions Uses standard network protocols The Double-Take solution has many useful features for an enterprise communication system, including: Continuous, asynchronous, byte-level replication Unlimited distance replication Hardware/application independent Certified Microsoft Windows compatibility Intelligent data compression Full voic redundancy as an option with DoubleTake More information on the Double-Take solution is available at Reliability The Way PAGE 13

14 4.2 VMware An alternative to Double-Take is VMware. The purpose of this solution is to allow customers to reduce their server footprint and therefore reduce costs by taking advantage of virtualization. Businesses can also leverage the high availability capability of VMware if they are looking for disaster recovery options. In the solution, the system management ( Director) software and headquarters software runs on the ESX hardware with version 4 software. Other solution components (telephony switches, Communicator, Call Center) run on their existing hardware components. software typically runs on a Microsoft Windows server operating system in a multiple VMware core environment. A generic depiction of the VMware solution can be seen in Figure 6. Figure 6. Generic VMware Solution for Reliability The Way PAGE 14

15 This figure does not show every configuration possible as the actual customer implementation depends upon the number of users on the system, reliability requirements (none, cold standby, or hot standby), the data storage required, and other equipment factors. The VMware solution offers the following options for customers: VMware support options No reliability High Availability (HA) this is a cold/warm restart with up to 3 minutes downtime vmotion allows users to move a running virtual machine from one server to another in real-time with no degradation of user capability Mass Storage Configurations Network Attached Storage NAS Storage Area Network SAN Customers are free to mix and match the options based upon the size of the virtual machines and dimensioning trade-offs that they wish to make. Features like vmotion allow system administrators to perform maintenance during working hours without having to schedule system downtime which can have definite productivity improvements. More information is available on VMware at Enterprise Contact Center Disaster Recovery Enterprise Contact Center has a built-in disaster recovery solution that ensures that your most critical call center functionality is available 24x7. The contact center application can run in a hot standby mode with automatic failover capabilities. This ensures that the customer experience is retained during failure conditions and the call center continues to generate revenue or improve customer satisfaction. advanced teleworker solutions also ensure that agents and supervisors continue to operate during network disruptions whether working at a remote site or teleworking from home. Agents can also login to the call center just by calling into an Interactive Voice Response (IVR) system if they lose their internet connection. Critical functions such as reporting is retained during any failure condition regardless of the location of the supervisor or the agent. Reliability The Way PAGE 15

16 5. Summary The UC system leads the industry in meeting the reliability and availability demands of businesses today. It combines powerful voice and UC features with a simple and easy-to-use management interface. This allows s customers to focus on what s truly important managing the business, not managing a communication system. customers have the flexibility to essentially create their own systems by choosing from the following needs: Distributed or centralized architecture The appropriate level of redundancy and cost The applications needed and the level of availability The level and type of disaster recovery needed By giving the customer choices, they have the flexibility to tailor a UC system to suit their individual business needs. For more information about products and solutions, please visit com or call (800) About is the provider of brilliantly simple Unified Communication (UC) solutions based on its award-winning IP business phone system. We offer organizations of all sizes integrated, voice, video, data, and mobile communications on an open, distributed IP architecture that helps significantly reduce the complexity and costs typically associated with other solutions. The feature-rich UC system offers the lowest total cost of ownership (TCO) and the highest customer satisfaction in the industry, in part because it is easy to deploy, manage, scale and use. Increasingly, companies around the world are finding a competitive edge by replacing business-as-usual with new thinking, and choosing to handle their integrated business communication. is based in Sunnyvale, California, and has regional offices and partners worldwide. For more information, visit shoretel.com. WORLD HEADQUARTERS 960 Stewart Drive, Sunnyvale, CA USA. EMEA ASIA PACIFIC +1 (800) Toll Free. +1 (408) Tel. +1 (408) Fax Freephone. +44 (1628) Tel. +61 (0) Tel. Copyright All rights reserved. The logo,, ShoreCare, ShoreGear, ShoreWare, ShorePhone and ControlPoint are registered trademarks of, Inc. in the United States and/or other countries. All other copyrights and trademarks herein are the property of their respective owners. Specifications are subject to change without notice. Part # /8.10